Guillaume Pernelle
Technical Leader and R&D scientist. Experienced in machine learning, neuroscience, medical imaging and finance. Successful exit founder.
Info
- gpernelle@gmail.com
- Lausanne, Switzerland
Summary
Accomplished professional with a robust blend of expertise in engineering, neuroscience, and machine learning. I hold a PhD from Imperial College London along with engineering degrees from École Centrale Marseille and Technische Universität München. As an entrepreneurial-minded leader, I have founded and successfully exited an AI startup, demonstrating my capability to innovate and drive technology ventures from conception to acquisition. I am highly skilled in leading research and development initiatives, seamlessly integrating the latest technologies into corporate environments, and possess a deep understanding of machine learning principles to ensure their swift and successful deployment. Additionally, my background includes effective management of distributed teams, which has consistently ensured project success and driven pioneering advancements.
Experience
- Founded Audiala Solutions AG, a company dedicated to pioneering new solutions for the tourism industry.
- Creating an innovative audio guide application, transforming travel experiences globally.
- Developed a solution to instantly translate Android and iOS applications and their store listings into any language.
- Led the implementation of large language models in the company's infrastructure, enabling the development of new products and services.
- Trained the team on the latest advances in machine learning and natural language processing, ensuring the successful deployment of new technologies.
- Created an AI chatbot and a new recommendation system.
- Participated in the Microsoft AI accelerator program.
Led the integration of Magic Carpet AI's technology into Blockchain.com's technology stack following the acquisition of our company.
Implemented improvements to Blockchain's infrastructure, including ease of deployment and procedures, reduced trading latencies by up to 50x.
Developed new data pipelines leveraging GCP BigQuery, BigTable, Airflow and other technologies to meet Blockchain's infrastructure constraints and to improve the efficiency and robustness of our data pipelines.
Conducted research on signal in orderbook and tick data to improve execution algorithms and develop new strategies.
MCAI - Deep Reinforcement Learning applied to Portfolio Management
Co-Founded Magic Carpet AI, a London-based startup focused on developing innovative investment strategies utilizing state-of-the-art machine learning concepts (deep reinforcement learning).
As CTO, led the research and development efforts, ensuring the focus was on the most rewarding tasks and maximizing the initial budget.
Developed a new family of stable and high-performing investment algorithms that learn from past conditions and adapt to current conditions to balance risks and returns.
Led the development of a complete in-house trading and risk management system, specifically tailored for digital assets.
Solutions for Testing and Quality Control
Experience as a software developer, building automated vision control solutions with data acquisition and signal and image processing.
Analyzed customer requirements and participated in designing technical solutions (optics, software, and hardware)
Developed software solutions (algorithms and GUI) and installed systems on customer premises
Responsible for testing, validating systems and creating technical documentation and user manuals
Led the development of QMTSubface, a project developed in partnership with the EPFL which uses machine learning to qualify the esthetic properties of objects.
Research Laboratory in the Department of Radiology of Brigham and Women's Hospital
Research Trainee in Radiology, Brigham and Women’s Hospital and Harvard Medical School. Finishing my study at TU München, I wrote my master thesis at the Surgical Planning Laboratory. This thesis spans mathematical algorithm development in image segmentation and registration using mechanically inspired models, design of an evaluation phantom and a dependent study, as well as efficient and well-documented software developed in Python and C++, which has been released publicly. This work has been followed by several peer-reviewed publications (MICCAI 2013, Medical Image Analysis 2017).
World’s First Minimally Invasive Non-blood Contacting Cardiac Assist Device to Treat Acute Heart Failure.
Prototyped analog hardware for reading an electrocardiogram (ECG) and driving a pump. This was part of the development of a non-blood contacting cardiac assist and support device system that pushes the heart wall synchronized with an ECG.
A340 Final Assembly Line
Intern on the final assembly line of the Airbus A340. Responsibilities included assembly of motors, winglets, and landing gear as well as on verification of electrical systems.
Ihre Software wie gewünscht
Part time job as a freelancer. Built a website for OptCon and performed the search engine optimization (SEO).
Talks and Workshops
Gap junction plasticity as a mechanism to regulate network-wide oscillations
Validation of catheter segmentation for MR-guided gynecologic cancer brachytherapy
Education
PhD in Computational Neuroscience
I modeled and analyzed the dynamics of neural networks, focusing on the role of gap junctions, or electrical synapses, in regulating the oscillations of electrical activity observed in many brain regions. My research showed that gap junction plasticity could play a crucial role in ensuring robust information transfer between brain regions. Additionally, I studied the relationship between blood oxygen concentration and neural activity. As a Graduate Teaching Assistant, I also assisted in Computational Neuroscience, Mathematics, and Python Programming courses.
Diplom Ingenieur (Masters)
- Major
- Medical Technology
- Minors
- Micro Technology, Product Development
I participated in the Top Industrial Managers for Europe program, which provides training for future leaders in the industry. During my degree, I completed my thesis at a lab affiliated with Harvard Medical School, where I applied research, software engineering, and mechanical engineering techniques to the medical field, specifically in radiology. This experience gave me a deep understanding of the latest advancements in the field, and the opportunity to collaborate with renowned researchers. It also allowed me to hone my research and analytical skills, as well as my ability to work effectively in a team.
Engineering Degree
- Major
- Image and Signal Processing
- Minors
- Management and Communication
Ingénieur Centralien (corresponding to Master) specialized in image processing and spatio-temporal signal processing. Thesis topic: Development of an imaging system based on light polarization for biomedical applications (detection of skin cancers).
Publications
Fully automatic catheter segmentation in MRI with 3D convolutional neural networks - application to MRI-guided gynecologic brachytherapy
Physics in Medicine & Biology 64 (16), 165008, 2019
Automatic needle segmentation and localization in MRI with 3-D convolutional neural networks - application to MRI-targeted prostate biopsy
IEEE transactions on medical imaging 38 (4), 1026-1036, 2018
Gap junction plasticity as a mechanism to regulate network-wide oscillations
PLoS computational biology 14 (3), e1006025, 2018
Accurate model-based segmentation of gynecologic brachytherapy catheter collections in MRI-images
Medical image analysis 42, 173-188, 2017
Gap junction plasticity can lead to spindle oscillations
arXiv preprint arXiv:1710.03999, 2017
Model-based catheter segmentation in MRI-images
arXiv preprint arXiv:1705.06712, 2017
Needle trajectory segmentation for mri-guided prostate biopsy
BWH Research Symposion, 2017
Modelling and analysing of neural network dynamics
Imperial College London, 2017
EM-navigated catheter placement for gynecologic brachytherapy - an accuracy study
Medical physics 44 (9), e270-e278, 2017
Validation of catheter segmentation for MR-guided gynecologic cancer brachytherapy
International Conference on Medical Image Computing and Computer-Assisted Intervention, 2013
Automated Interstitial Brachytherapy Catheter Localization From Volumetric MR Data
Medical Physics 40 (6Part2), 94-94, 2013
Robust Applicator Registration for Interstitial Gynecologic Brachytherapy
Brachytherapy 12, S53, 2013
Needle Labeling for Interstitial Gynecological Brachytherapy
5th NCIGT and NIH Image Guided Therapy Workshop. NCIGT, NAC, Boston 48, 2012
Segmentation of pelvic structures for gynecologic brachytherapy
5th NCIGT and NIH Image Guided Therapy Workshop, 45, 2012
Development of a Non-Blood Contacting Cardiac Assist and Support Device - An In Vivo Proof of Concept Study
Journal of Medical Devices 5 (4), 041007, 2011
Assessment of Minimally Invasive Device That Provides Simultaneous Adjustable Cardiac Support and Active Synchronous Assist in an Acute Heart Failure Model.
Journal of Medical Devices 5 (4), 041008,, 2011
Open Source
Deep Learning applied to Needle Segmentation
Convolutional neural networks (3D U-Net) model to segment needles from 3D MRI, using the Tensorflow framework.
NeedleFinder: fast interactive needle detection.
Open-source software extension that provides interactive tools to segment needles in MR/CT images. It has been mostly tested on MRI from gynelogical brachytherapy cases. Cf <
> MICCAI 2013 iGyne: Tools to for Image Guided Brachytherapy
iGyne is currently articulated in seven steps:
- procedure selection
- applicator selection
- data importation
- initial applicator registration
- refined applicator registration
- needle position planning
- needle detection